Hymba 1.5B Instruct — Hardware Requirements & GPU Compatibility
Chat
454 downloads
243 likesSpecifications
- Publisher
- NVIDIA
- Parameters
- 1.5B
- Architecture
- HymbaForCausalLM
- Context Length
- 8,192 tokens
- Vocabulary Size
- 32,001
- Release Date
- 2025-01-02
- License
- Other
Get Started
HuggingFace
How Much VRAM Does Hymba 1.5B Instruct Need?
Select a quantization to see compatible GPUs below.
| Quantization | Bits | VRAM | + Context | File Size | Quality |
|---|---|---|---|---|---|
| BF16 | 16.00 | 3.4 GB | 3.7 GB | 3.05 GB | Brain floating point 16 — preferred for training |
Which GPUs Can Run Hymba 1.5B Instruct?
BF16 · 3.4 GBHymba 1.5B Instruct (BF16) requires 3.4 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 5+ GB is recommended. Using the full 8K context window can add up to 0.3 GB, bringing total usage to 3.7 GB. 35 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti.
Runs great
— Plenty of headroomWhich Devices Can Run Hymba 1.5B Instruct?
BF16 · 3.4 GB33 devices with unified memory can run Hymba 1.5B Instruct, including NVIDIA DGX H100, NVIDIA DGX A100 640GB.
Runs great
— Plenty of headroomRelated Models
Frequently Asked Questions
- How much VRAM does Hymba 1.5B Instruct need?
Hymba 1.5B Instruct requires 3.4 GB of VRAM at BF16.
VRAM = Weights + KV Cache + Overhead
Weights = 1.5B × 16 bits ÷ 8 = 3 GB
KV Cache + Overhead ≈ 0.4 GB (at 2K context + ~0.3 GB framework)
KV Cache + Overhead ≈ 0.7 GB (at full 8K context)
VRAM usage by quantization
BF163.4 GBBF16 + full context3.7 GB- Can I run Hymba 1.5B Instruct on a Mac?
Hymba 1.5B Instruct requires at least 3.4 GB at BF16, which exceeds the unified memory of most consumer Macs. You would need a Mac Studio or Mac Pro with a high-memory configuration.
- Can I run Hymba 1.5B Instruct locally?
Yes — Hymba 1.5B Instruct can run locally on consumer hardware. At BF16 quantization it needs 3.4 GB of VRAM. Popular tools include Ollama, LM Studio, and llama.cpp.
- How fast is Hymba 1.5B Instruct?
At BF16, Hymba 1.5B Instruct can reach ~850 tok/s on AMD Instinct MI300X. On NVIDIA GeForce RTX 4090: ~191 tok/s. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = (bandwidth GB/s ÷ model GB) × efficiency
Example: AMD Instinct MI300X → 5300 ÷ 3.4 × 0.55 = ~850 tok/s
Estimated speed at BF16 (3.4 GB)
AMD Instinct MI300X~850 tok/sNVIDIA GeForce RTX 4090~191 tok/sNVIDIA H100 SXM~635 tok/sAMD Instinct MI250X~525 tok/sReal-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
- What's the download size of Hymba 1.5B Instruct?
At BF16, the download is about 3.05 GB.
- Which GPUs can run Hymba 1.5B Instruct?
35 consumer GPUs can run Hymba 1.5B Instruct at BF16 (3.4 GB). Top options include AMD Radeon RX 6700 XT, AMD Radeon RX 6800, AMD Radeon RX 6800 XT. 35 GPUs have plenty of headroom for comfortable inference.
- Which devices can run Hymba 1.5B Instruct?
33 devices with unified memory can run Hymba 1.5B Instruct at BF16 (3.4 GB), including Mac Mini M4 (16 GB), Mac Mini M4 (32 GB), Mac Mini M4 Pro (24 GB), Mac Mini M4 Pro (48 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.